[MPlayer-users] quant levels? pass.log xvid.dbg?

jzaw jz at w.dnsalias.org
Thu Nov 27 16:52:24 CET 2003


hi all ... ive been an avid reader of the list for a while .. thanks  
for the fantastic info everyone gives
now my first question  --- SORRY its such a long one

using
MEncoder 1.0pre2-3.3
lame-3.93.1
xvidcore-0.9.2
all compiled within the last month

im making 2 pass xvid's of my dvd collection for a personal video juke  
box
im coding to 700MB for cd archiving purposes so compresion on longer  
films is greater
im hitting the 700MB +- 500k mark every time but as was discussed on a  
previous thread the xy scaling can make a vast difference to the  
playback quality ... bit per pixel being the important factor discussed

ive incorporated important elements of calcbpp.pl (many thanks to the  
author) in to my own script to give me a table of xy scaling with bpp  
values so i can make the choice

further ive tried to use countquant.pl to give me quant stats from the  
pass log so i can follow the advice in the encoding tips txt and select  
an xy that gives me mostly quant 2 values but it errors saying there  
are no quant values

HOWEVER

looking at the pass log ... all i get is quant 2 ... regardless of any  
xy settings what so ever ... even from 320x144 to 720x304 keeping all  
other settings the same
it doesnt make a difference i still get all quant 2 values in the pass  
log

not till i enabled xvid debug

> <from man file>
> xvidenc (-xvidencopts) debug

did i see any values other than quant 2 ... it was shocking !
as you can see all the quants are very very high .... (are they high?)

am i now finally looking at significant info?
ie was the pass log irrelevant and didnt contain info that was coherent  
in any way
and xvid.dbg is giving me true info that i can use to make informed  
scaling decisions?

as per encoding tips txt does this mean i can increase my xy size?
(the logs for test run shown below was at 640x272)
the suggestion being that making sure there are few quant 3 and quant  
4's present and mostly quant 2's

btw for this test film the bpp at 640x272 is 0.14
lower than the encoding tips state as a candidate for a value but  
overall ive found that 0.2 - 0.13 works well on most scifi films you  
know .. the dark of space few speckles for stars etc not much bitrate  
required (visual inspection confirms)

further im finding it difficult to see any difference or improvement in  
playback between using 4mv (at encode time) and not using it.
using it seems to only add about 20 mins to the current 4 - 5  hours  
taken for a transcode and it doesnt decrease the file size (of course)


command used in a 2 pass loop for this particular film

/usr/local/bin/mencoder dvd://1 -dvd-device /pathto/FILM_NAME/VIDEO_TS  
-dvdangle 1 -oac mp3lame -lameopts cbr:br=128:vol=3 -aid 128 -ovc xvid  
-sws 2 -xvidencopts  
bitrate=610:max_key_interval=240:4mv:debug:pass=$pass -passlogfile  
/pathto/FILM_NAME_transcode.log -ofps 25 -vop  
scale=640:272,crop=718:436:2:70 -o /pathto/FILM_NAME_xvid.avi


all comments welcome
any comments re b frames?

thanks in advance

Zaw

SNIPPETS  OF BPP TABLE AND  BOTH LOGS FOLLOW

xy scal bpp table

> Prescaled picture: 1021x436, AR 2.35 (2.34 uncorrected)
> 720x304, diff   2, new AR 2.37, AR error 0.78% scale=720:304 bpp: 0.111
> 704x304, diff  -4, new AR 2.32, AR error 1.48% scale=704:304 bpp: 0.114
> 688x288, diff   5, new AR 2.39, AR error 1.63% scale=688:288 bpp: 0.123
> 672x288, diff  -2, new AR 2.33, AR error 0.71% scale=672:288 bpp: 0.126
> 656x272, diff   7, new AR 2.41, AR error 2.56% scale=656:272 bpp: 0.137
> 640x272, diff   0, new AR 2.35, AR error 0.12% scale=640:272 bpp: 0.140
> 624x272, diff  -6, new AR 2.29, AR error 2.44% scale=624:272 bpp: 0.144
> 608x256, diff   3, new AR 2.38, AR error 1.05% scale=608:256 bpp: 0.157
> 592x256, diff  -4, new AR 2.31, AR error 1.62% scale=592:256 bpp: 0.161
> 576x240, diff   5, new AR 2.40, AR error 2.08% scale=576:240 bpp: 0.177
> 560x240, diff  -2, new AR 2.33, AR error 0.71% scale=560:240 bpp: 0.182
> 544x224, diff   7, new AR 2.43, AR error 3.24% scale=544:224 bpp: 0.200
> 528x224, diff   1, new AR 2.36, AR error 0.30% scale=528:224 bpp: 0.206
> 512x224, diff  -6, new AR 2.29, AR error 2.81% scale=512:224 bpp: 0.213
> 496x208, diff   3, new AR 2.38, AR error 1.45% scale=496:208 bpp: 0.237
<snip>



xvid.dbg output

> # XviD Debug output
> # quant | intra | header bytes| total bytes | kblocks | mblocks |  
> ublocks| vbr overflow | vbr kf overflow| vbr kf partial overflow
>
> 6 1 1897 1897 680 0 0 0 0 0
> 4 0 92 92 0 0 680 0 -1312 -5
> 2 0 687 2387 0 680 0 -5 -1307 -5
> 2 0 92 92 0 0 680 -2305 -1302 -5
> 2 0 92 92 0 0 680 -2310 -1297 -5
> 2 0 92 92 0 0 680 -2315 -1292 -5
> 2 0 92 92 0 0 680 -2320 -1287 -5
> 2 0 92 92 0 0 680 -2325 -1282 -5
<snip>
> 3 0 92 92 0 0 680 -2470 -1137 -5
> 2 0 92 92 0 0 680 -2475 -1132 -5
> 2 0 92 92 0 0 680 -2480 -1127 -5
> 2 0 92 92 0 0 680 -2485 -1122 -5
> 3 0 115 128 0 22 658 -2490 -1117 -5
> 3 0 310 524 0 218 462 -2410 -1112 -5
> 5 0 482 855 62 145 473 -2442 -1107 -5
> 6 0 962 1604 92 404 184 -2305 -1102 -5
> 6 0 1344 2246 86 514 80 -2312 -1097 -5
> 7 0 1736 2898 158 476 46 -2418 -1092 -5
> 7 0 1414 2365 46 596 38 -2540 -1087 -5
> 6 0 1197 2254 3 642 35 -2389 -1082 -5
> 7 0 1270 2172 24 606 50 -2479 -1077 -5
> 7 0 1210 2133 5 638 37 -2343 -1072 -5
> 6 0 1375 2970 15 637 28 -2110 -1067 -5
> 7 0 1144 1983 5 641 34 -2590 -1062 -5
> 7 0 1225 2435 13 635 32 -2210 -1057 -5
> 7 0 1263 2581 13 637 30 -2054 -1052 -5
> 6 0 1396 3688 10 664 6 -1994 -1047 -5
> 7 0 1273 2675 11 657 12 -2740 -1042 -5
> 7 0 1274 3087 7 646 27 -2434 -1037 -5
> 7 0 1217 2871 5 646 29 -2352 -1032 -5
> 7 0 1308 3846 10 648 22 -2268 -1027 -5
> 7 0 1296 3814 7 664 9 -2321 -1022 -5
<snip>
> ####################################################################### 
> ########
> # Quantizer distribution : xvid
>
> # quant 1 : 0
> # quant 2 : 568
> # quant 3 : 8931
> # quant 4 : 60636
> # quant 5 : 69150
> # quant 6 : 43687
> # quant 7 : 10997
> # quant 8 : 1756
> # quant 9 : 338
> # quant 10 : 22
> # quant 11 : 4
> # quant 12 : 2
> # quant 13 : 0
> # quant 14 : 0
> # quant 15 : 0
<snip>
<end>





pass log output : LOOK ALL QUANT 2 !!!

> # ASCII XviD vbr stat file version 0
> #
> # frames    : 0000196091
> # keyframes : 0000002700
> #
> # quant | intra | header bytes | total bytes | kblocks | mblocks |  
> ublocks
>
> 2 1 1897 1897 680 0 0
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
<snip>
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 92 92 0 0 680
> 2 0 165 213 0 73 607
> 2 0 346 669 0 251 429
> 2 0 789 2284 62 416 202
> 2 0 1240 4336 72 577 31
> 2 0 1534 6277 49 587 44
> 2 0 1980 8697 134 533 13
> 2 0 1666 7685 27 632 21
> 2 0 1491 6368 0 662 18
> 2 0 1491 6904 6 656 18
> 2 0 1482 10218 6 674 0
> 2 0 1447 9417 5 675 0
> 2 0 1472 12495 6 674 0
> 2 0 1501 12870 4 676 0
> 2 0 1283 9464 1 679 0
> 2 0 1249 11524 0 679 1
> 2 0 1191 10034 0 679 1
> 2 0 1141 9619 0 678 2
> 2 0 1094 10351 0 675 5
<snip>
> 2 0 778 8041 0 599 81
> 2 0 648 6883 0 501 179
> 2 0 569 5361 0 461 219
> 2 0 614 4837 0 495 185
> 2 0 605 4790 0 481 199
> 2 0 585 5653 0 470 210
> 2 0 601 6124 0 485 195
> 2 0 565 5015 0 460 220
<end>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 7301 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-users/attachments/20031127/f30153f4/attachment.bin>


More information about the MPlayer-users mailing list